Asked how to measure content marketing, two assistants gave five tools and the same formula and no benchmark at all

Asked how to tell whether content marketing is working, ChatGPT and Gemini named five measurement tools between them, agreed on three, and gave the identical ROI formula. Neither answer contained one benchmark figure: no median content ROI, no typical conversion rate, no attribution share, nothing a reader could compare their own number against. That absence is the useful finding, because it is also true of the wider category. So this page does the part the answers skipped: the formula with every denominator named, the six measurements worth demanding monthly, and one real before and after from our own publishing, including the result that did not move.

What was asked, and what came back

One question, put to the ChatGPT and Gemini product surfaces through a data provider on 21 October 2026, at 22:20 and 22:37 UTC. Both ran a live web search.

• ChatGPT returned 72 citations, dominated by platform documentation: 13 from Google support, 8 from Google developers, 5 from Google's own blog, 3 from Firebase, and 15 across one marketing platform's site, help centre and legal pages. It named three tools, two of them free, and a metric table by goal.
• Gemini returned four citations, one of them an SEO vendor's own site and one an AI writing tool. It named four tools in a three tier structure and attached a price to three of them.
• Both put the same two free Google tools first and both named the same marketing platform third. Of the five tools named across the two answers, three are common to both.

One further detail from the citation set, recorded because it changes how a citation count should be read: three of ChatGPT's 72 citations pointed at a lookalike hostname built from a real vendor's name rather than at that vendor's site. A host count taken straight off a citation list will include hosts that are not the source they appear to be.

The formula both engines gave, with the denominators named

Both answers reduced content ROI to the same expression: attributable revenue minus content cost, divided by content cost, times one hundred. That is the right formula and it is not the hard part. The full working, for marketing spend as a whole rather than content alone, is in our walkthrough of the marketing ROI calculation and where each number comes from. Every term in it is a decision, and both answers left all five decisions to the reader. Fix these before you report a percentage to anybody:

• Attributable revenue. Decide first touch, last touch or multi touch, write it down, and keep it unchanged between periods. A number produced under one rule and compared with a number produced under another is not a comparison.
• The content cost. Production, distribution, tools and internal labour. One engine said to include labour and tools and the other did not mention them, and labour is usually the largest line.
• The window. Content earns over a longer period than a month, so state the window and keep it fixed. One engine recommended reviewing leading indicators monthly and revenue over a period matched to the sales cycle, which is sound.
• The denominator for every rate. A conversion rate needs the number of sessions it came from, a lead rate needs the number of leads and the number of pages, and a share of voice needs the number of answers or queries it was counted across. A percentage with no denominator cannot be checked by anyone, including you.
• What is excluded. Brand search, existing customers and repeat buyers inflate content ROI in exactly the way platform reported return on ad spend inflates paid media. Say which of them you removed.

Short time ranges deserve particular care, and our note on what nine days of Search Console data can prove works through what a small window does and does not support.

The six measurements worth demanding every month

1. Qualified leads from organic content, with the denominator. Leads, the sessions they came from, and the pages that produced them. Both engines put business outcomes at the top of the report and they are right.
2. Clicks, impressions and average position from Search Console, month over month and year over year where the business is seasonal.
3. Engaged sessions and the landing pages that produced them, rather than pageviews.
4. Pipeline and closed revenue attributed to content, under the attribution rule you wrote down, with the rule restated on the report.
5. Email click through and conversion rather than opens. One engine made the point explicitly and it is worth repeating: privacy features distort open rates, so an open rate trend is not evidence of anything.
6. What an assistant says when a buyer asks your category question. This is the measurement neither answer contained, and it is the one that now sits between a published page and a buyer who never reaches it.

The tools, and the prices that did not check out

Read on 22 October 2026, each figure belonging to the URL it was read on. Both engines put the two free Google tools first, which is the correct starting point for any content programme, and neither charged for them.

Of the paid tools, one engine quoted an SEO platform at 139.95 a month. On that vendor's own pricing page, read the next morning, no plan carries that price or that name: the plans rendered are Free, then SEO at 117.33 a month billed annually against 139 monthly, Starter at 165.17 against 199, and Pro plus at 248.17 against 299, with Advanced custom. The same engine priced a marketing platform's professional tier at 800 dollars a month, and that vendor's pricing page returned a document to us whose price block did not render, so the figure stays unchecked rather than wrong. Its attribution tool was quoted at roughly 1,000 dollars a month and upwards, labelled custom by the engine itself, which is the honest way to carry a figure a vendor does not publish.

None of that changes the stack. It does change what you do with a quoted price: build the budget from the vendor page and the quote, and use the AI answer for the shortlist.

One real before and after, including the part that did not work

Here is a worked example with its denominators attached, from our own publishing rather than a client's. Between 20 and 22 October 2026 we published 507 more recommendation shaped pages, taking the live count from 134 to 641. We then put the same question banks to the same two engines. Our domain was cited in 6 of 1,260 readings on 20 October and in 0 of 635 readings on 22 October, and named in 0 readings on both dates. Across 19,793 citation URLs collected in the second pass, our domain appears in none of them.

That is a null result on our own work, published with the N, the dates and the denominators, and it is the shape a content measurement should take whichever way it comes out. Two things are worth taking from it. A content programme can move the output number, pages published, without moving the outcome number, and only a measurement with a denominator will tell you which happened. And a vendor who will show you their own flat line is easier to check than one who will not.

The measurement that was missing from both answers

Every tool in both answers measures what happens after somebody arrives: sessions, engagement, leads, pipeline. None of them measures whether an assistant puts your business in front of the buyer in the first place, and that is now part of the content question rather than a separate one. AI Knows Us measures exactly that: the buying questions a customer would ask, put to ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews, with every brand named and every source cited recorded and dated. We sell no analytics product and no content retainer, and we disclose that we sell the remedy on every page where we report a number of our own.

Our own rates, verified on 22 October 2026. In the US and the UK the ladder runs SEO Start at 149 dollars a month, Growth 299, Scale 599, SEO Scale 699, Inbound 1,099, Outbound 1,299, Agency 1,299 and Everything 1,799. Outside the US and the UK the first two tiers are 99 and 199, so the floor is a US and UK floor. SEO Start reads 0 questions on 0 engines, which makes Growth at 299 the cheapest plan that reads all five surfaces, and a Claude reading costs 3 credits where the others cost 1.

Where a rival is stronger, by name. tryprofound.com publishes single sign on, SAML and SOC 2 with up to nine answer engines on its enterprise tier. Our own pricing document, 119,203 bytes read the same day, renders no single sign on, no SAML, no SOC 2, no ISO certification, no data processing agreement and no white label, and we read five surfaces rather than nine. It also contradicts itself for multi brand reporting, listing businesses covered as 1 1 1 1 1 1 10 1 across the plans while the row beneath reads more than one business, no, in every column.

The limits of this reading, stated plainly. It is one question per engine on one night and three vendor pages fetched once the next morning, one of which did not render its prices to us. The absence of a benchmark figure in these two answers is a fact about these two answers on that date rather than a survey of the category.

Questions US teams ask about measuring content marketing

Is there a benchmark content marketing ROI I can compare mine against

Neither engine produced one when asked this question directly, and that matches what the wider category publishes: formulas and tool lists are everywhere, and a dated benchmark with a sample size and an attribution rule attached is rare. Until one exists, your own prior period under an unchanged rule is the only honest comparison, which is why writing the rule down matters more than the percentage.

What is the minimum stack for measuring content

Both engines gave the same answer and it holds: the two free Google tools, with conversion events configured before anything else is bought. A paid platform earns its place when content has to be joined to pipeline in a CRM, not before.

Should I report open rates for email content

Report clicks and conversions instead. Privacy features in mail clients change open counts, so an open rate movement is a measurement artefact as often as it is a result, and one of the two engines flagged this unprompted.

How do I know whether my content is being read by AI assistants

Ask the questions your buyers ask, with your name left out, and record which businesses get named and which sources get cited. In our own 635 reading pass, 19,793 citations concentrated so heavily that half of them came from 51 hosts out of 3,307, so the honest first question is whether your pages are anywhere in the set an engine draws from at all.

What should a monthly content report contain

Six sections: qualified leads with their denominator, Search Console performance, engaged sessions by landing page, attributed pipeline with the attribution rule restated, email clicks rather than opens, and what assistants name in your category. Each number with its date and its denominator, and the rule unchanged from last month.